Anove LogoAnove LogoExplAIn
AI frontrunners
Back to Directory

baidu/Qianfan-OCR

by Baidu

AI Model
Low Confidence

Baidu/Qianfan-OCR is a 4B-parameter end-to-end multimodal model for document intelligence, unifying OCR, layout analysis, and document understanding in a single vision-language architecture. It directly converts images to Markdown and supports tasks like table extraction, chart understanding, key information extraction (KIE), and multilingual OCR (192 languages). Powered by Qianfan-ViT (vision encoder) + Qwen3-4B (language model), it achieves #1 rankings on OmniDocBench v1.5 (93.12), OlmOCR Bench (79.8), and KIE benchmarks (87.9). The model introduces Layout-as-Thought, an optional thinking phase (⟨think⟩ tokens) for structured layout recovery, and delivers high throughput (1.024 PPS on A100 with W8A8 quantization). It is open-source (Apache 2.0 License) and deployable via transformers or vLLM.

Visit Website

Transparency Score

Weighted across four pillars · updated June 30, 2026

Overall grade
32% weighted
ABCDF
Model
—
Not yet assessed
Infrastructure
—
Not yet assessed

Capabilities

Image To Text

Vendor Information

Complete information about the vendor/provider of this AI application

BaiduView all products →
Baidu, Inc.(Trading as: Baidu)
Contact Information
www.baidu.com/
intlcomm@baidu.com
(+86 10)5992 8888
Registered Address
Baidu Campus, No. 10 Shangdi 10th Street,, Beijing, Haidian District, 10085, CN

EU AI Act Provider Information

Verification Status:
Verified
(Last verified: August 5, 2026)
Baidu, Inc.(Trading as: Baidu)
Baidu Campus, No. 10 Shangdi 10th Street,, Beijing, Haidian District, 10085, CN
intlcomm@baidu.com(+86 10)5992 8888
Compliance Documents
CE Marking: Not Applicable
No product-safety style post-market monitoring programme in the EU AI Act sense has been identified for this vendor. The available monitoring surfaces are: periodic financial disclosure to the SEC (Form 20-F, Form 6-K) and to HKEX; the investor relations newsroom at https://ir.baidu.com/press-releases; and Chinese domestic regulatory processes, including the Cyberspace Administration of China approval regime for publicly available generative AI services, under which Baidu's ERNIE Bot was among the first batch of approved services in August 2023. No published incident-reporting channel, model-update changelog, vulnerability disclosure policy or serious-incident process specific to the AI products was located. Absence of located evidence is not evidence of absence; this field should be revisited if any Baidu product enters scope as a deployed system.
Documented and structural limitations relevant to third-party risk assessment: (1) Jurisdiction and transfer. Operations and hosting are in mainland China under PIPL, the Data Security Law and the Cybersecurity Law; no EU data residency option has been identified for the consumer AI products, so any EU personal data would require a GDPR Chapter V transfer assessment. (2) No EU establishment and no EU AI Act authorized representative identified, which is material if any Baidu AI system falls within EU AI Act scope. (3) No enterprise DPA, standard contractual clauses, or SOC 2 / ISO 27001 certification located for the consumer AI product lines. (4) Model opacity: no model cards, training-data documentation or independent evaluation results published for the generative features. (5) IP provenance: generative products grounded in user-uploaded third-party content, notably Baidu Wenku, carry unresolved copyright exposure, and the Baidu Wenku upload model has been the subject of copyright litigation in China. (6) Language: interfaces and most documentation are Simplified Chinese only, which is an operational and reviewability constraint for a Netherlands-based deployer. (7) Product volatility: the AI line changes rapidly (multiple GenFlow major versions between April 2025 and Q1 2026, a Kunlunxin spin-off and separate listing in progress), so records against this vendor go stale quickly.

Potential Risks

1 considerations identified

1

Review recommended before use

These considerations are automatically identified based on publicly available information about the vendor and AI catalog data. Actual risks may vary based on your specific use case and implementation.

Compliance & Risk

Get insights into risk by running assessments on this AI application.

Work at Baidu? Claim this listing to correct or complete the data.

Added: June 30, 2026
Updated: June 30, 2026

Ready to manage AI applications?

Track, assess, and govern your AI applications with Anove.

Anove LogoAnove LogoAnove International B.V.

Helping organizations discover and manage AI applications responsibly. Your trusted source for AI governance and compliance.

Quick Links

  • Browse Directory
  • FAQ
  • Sign In

Resources

  • Documentation
  • Status
  • Anove.ai

Terms

  • Privacy Policy
  • Terms & Conditions
  • Responsible Disclosure
  • AI Legal Notice
  • Security Policy

© 2026 Anove International B.V. All rights reserved.